Awesome GitHub RepositoriesDistributed Debugging

Ray is a distributed computing framework designed to scale Python and Java applications across clusters by abstracting task scheduling and resource management. It functions as a resource-aware execution engine that manages task dependencies, placement, and fault tolerance across networked compute nodes. At its core, the system provides a stateful actor model, allowing developers to define classes that run in dedicated processes to maintain and mutate internal state across remote method calls. The framework distinguishes itself through a robust cross-language interoperability layer, enabling f

This project is a distributed training infrastructure designed for aligning large language models through reinforcement learning. It functions as an end-to-end engine for complex alignment tasks, including proximal policy optimization, direct preference optimization, and iterative self-play. By providing a unified framework for multi-turn interactions and tool-use scenarios, it enables the development of models capable of reasoning and external environment engagement. The framework distinguishes itself through a decoupled architecture that separates model training from sample generation. This

Temporal is a distributed workflow orchestration engine designed to manage fault-tolerant, stateful, and long-running background processes. It functions as a platform for coordinating complex cross-service operations, ensuring consistency and reliability in distributed environments by decoupling workflow orchestration from task execution. The platform distinguishes itself through a deterministic, event-sourced execution model that reconstructs workflow state by re-executing code from an immutable event log. This approach isolates non-deterministic side effects into managed activities, allowin

Dask ist ein Framework für paralleles Rechnen und ein verteilter Task-Scheduler, der darauf ausgelegt ist, Python-Data-Science-Workflows von einzelnen Maschinen auf große Cluster zu skalieren. Es fungiert als Cluster-Ressourcenmanager, der die Berechnungslogik orchestriert, indem Aufgaben und deren Abhängigkeiten als gerichtete azyklische Graphen dargestellt werden. Diese Architektur ermöglicht es dem System, die Verteilung von Workloads auf verfügbare Hardware zu automatisieren und gleichzeitig komplexe Ausführungsanforderungen zu verwalten. Das Projekt zeichnet sich durch eine Lazy-Evaluation-Engine aus, die Datenoperationen verzögert, bis sie explizit angefordert werden, was eine globale Graphoptimierung und effiziente Ressourcenzuweisung ermöglicht. Es integriert speicherbewusstes Data-Spilling, um Systemabstürze bei der Verarbeitung von Datensätzen zu verhindern, die den verfügbaren Speicher überschreiten, und nutzt Task-Graph-Fusion, um Sequenzen von Operationen in einzelne Ausführungsschritte zu kombinieren, wodurch Scheduling-Overhead und Inter-Node-Kommunikation minimiert werden. Die Plattform bietet eine umfassende Oberfläche für die Datenanalyse im großen Maßstab, einschließlich Unterstützung für verteiltes maschinelles Lernen, Integration in das Hochleistungsrechnen und parallele Datenverarbeitung. Sie bietet umfangreiche Werkzeuge für das Cluster-Lebenszyklusmanagement, Performance-Profiling und die Echtzeitüberwachung der Aufgabenausführung. Benutzer können diese Umgebungen über verschiedene Infrastrukturen hinweg bereitstellen, einschließlich lokaler Hardware, Cloud-Anbietern, containerisierten Systemen und Hochleistungsrechner-Clustern.

Cadence is a distributed workflow orchestration engine designed to execute long-running, asynchronous business logic with built-in durability and resilience across distributed systems. It functions as a stateful process manager that ensures processes resume from their last known state following system crashes or network outages. The platform utilizes a distributed task queue to manage work across independent worker nodes and supports persistence via SQL or Cassandra backend storage. It includes a workflow visualization dashboard for inspecting execution histories and state traces, alongside a

BytePS is a distributed deep neural network training framework and communication library designed to scale model training across multiple GPUs and compute nodes. It functions as a GPU cluster orchestrator and RDMA network optimizer, providing the necessary primitives to synchronize gradients and data across a server cluster. The project distinguishes itself through high-performance network optimizations, utilizing remote direct memory access and page-aligned memory to reduce latency. It employs topology-aware communication tuning and CPU core affinity management to maximize hardware throughpu